搜索优化
English
搜索
Copilot
图片
视频
地图
资讯
购物
更多
航班
旅游
酒店
笔记本
Top stories
Sports
U.S.
Local
World
Science
Technology
Entertainment
Business
More
Politics
时间不限
过去 1 小时
过去 24 小时
过去 7 天
过去 30 天
按相关度排序
按时间排序
18 天
提升神经网络鲁棒性:基于权重分摊的LeNet-5防御创新
随着人工智能技术在自动驾驶、医疗诊断等重要领域的层出不穷应用,保障神经网络的鲁棒性与安全性已成为科学界的热门话题与挑战。当面对一系列潜在威胁,如对抗攻击、数据中毒及后门攻击等,随机翻转攻击尤其值得警惕,因为它能通过内置权重参数的微调直接影响到网络性能,危及其安全。 为了应对这一挑战,研究者们提出了一种名为基于权重分摊的防御策略。该策略巧妙地通过计算与分析权重的梯度来识别出关键神经元,并为这些神经元 ...
一些您可能无法访问的结果已被隐去。
显示无法访问的结果
今日热点
Los Angeles wildfire updates
Sanctions against Columbia
Confirmed as DHS secretary
38 hurt on United flight
Flights for Afghans paused
NK tested cruise missiles?
Shifts view on COVID origins
Escaped monkeys captured
Indicted for insurrection
Assault trial begins
‘Walk It Out’ rapper dies
Confirmed to lead Pentagon
Claims 1st win of the season
Smuggling deaths guilty plea
Win sixth ice dance title
Proposed ban withdrawn
US has not stopped aid
IA immigration law blocked
Recalls over 80,000 SUVs
Jets hire Mougey as GM
Hack impacted 190M
Sudan: At least 70 killed
DOJ drops case
Wins first Grand Slam title
Wins US figure skating title
Sentenced to 17+ years
153 war detainees freed
Carroll to coach Raiders
ISR to get 2K-pound bombs
PETA activists arrested
US lifts bomb restrictions
Nets retire No. 15 jersey
Graham on Jan. 6 pardons
反馈